Use of Placebos in Controlled Trials

Background The traditional ‘double-blind’ RCT

uses a placebo to conceal allocation.

There are a number of advantages to using a placebo that reduces non-selection bias.

Forms of Placebos A placebo should look feel and

taste the same as the active treatment, sometimes this is difficult.

Sham surgery is sometimes used as a ‘placebo’ surgery.

Placebo ‘talk’ therapy in psychological therapies.

Reduction in Subversion Bias A placebo ‘blinds’ or masks the following:

Doctor Patient Assessor

This reduces the possibility of subversion as the person recruiting the patient is ‘blind’ to treatment allocation and therefore cannot subvert. HOWEVER, drug codes can be cracked and subversion therefore can occur.

For example Sometimes placebos come in

different bottles or labelled differently if someone manages to crack the code then all allocations are unmasked.

Hawthorne Effect and Resentful Demoralization

Because patients do not know if they are on active treatment or not they will not be demoralised if they do not get active treatment.

All patients are taking a similar looking treatment and therefore they should all have the Hawthorne effect.

Ascertainment Bias Because patients are on placebos

they are not likely to differentially report events due to knowledge of their treatment. This means any differences can be more likely ascribed to treatment rather ascertainment bias.

Also assessors are blinded because of placebos.

Explanatory Effects Because placebos eliminate bias

they are more likely to produce a ‘truer’ estimate of a therapeutic effect than an ‘open’ trial.

There are however, problems through using placebo control.

Technical Problems It is not UNKNOWN for placebos to

actually be an active treatment Mistakenly at least one trial

labelled active drugs as placebo so ALL patients got active treatment – showing no effect.

Placebos can be unmasked.

Active or Inactive placebos Sometimes a placebo may not be totally

‘inert’. For example, some wound care trials

have used ‘placebo’ gel, which may actually have had a beneficial or harmful effect on the wound.

Often trials used ‘active’ placebos as it is unethical to allocate an inactive placebo. May be difficult to develop placebo.

Placebos Unmasked For some treatments placebo control

is difficult as active treatment produces an effect which immediately unblinds the person taking it (and/or their clinician).

This process can reduce the usefulness of a placebo as the unmasked patients can differentially respond.

Surgical Placebos Sham or placebo surgery is

sometimes used to mask the patient to their treatment (hopefully surgeon hasn’t been blinded).

Many think this is unethical as the control patients are exposed to harm (e.g. infection risk, anaesthesia) with no possibility of benefit.

Drawbacks of Placebos For many treatments a placebo

cannot be used (e.g most surgery). Even when a placebo can be used

it may not produce the most useful answer.

It may be better NOT to use a placebo.

Efficacy or Effectiveness? In routine clinical practice placebos

are not used. The EFFECT of a treatment is the

EFFICACY of the biological effect of treatment plus any benefit due to the ‘placebo’ effect.

Using placebos efficacy is estimated NOT effectiveness.

Trial of Antibiotics Little and colleagues undertook an

RCT of antibiotics for sore throat. They could have used placebos but

chose instead to use an ‘open’ design.

In the initial study it was shown that the use of antibiotics had NO effect on sore throat.

Number of patients returning for more

treatment

0

5

10

15

20

25

30

35

40

Antibiotic None

Little et al. BMJ 1997;315:350-52.

Antibiotics In the Little study it was shown that

participants randomised to antibiotics mistakenly attributed their improvement to treatment and therefore were more likely to re-attend their GP during the next episode of sore throat.

A placebo trial would NOT have found this result.

Placebos and Recruitment Gossip (i.e., qualitative data)

suggests that recruitment may be better in a non-placebo trial.

More robustly an RCT of placebos has indicated participants are more willing to take part in an ‘open’ rather than a placebo trial.

The ‘OPEN’ Trial An RCT of randomising men and

women (70+) to be asked whether or not they would take part in a fracture prevention trial examined the question of trial recruitment.

The aim was to test the hypothesis that open trials recruited better than placebo studies.

Diagram of OPEN trial

R e fusedT o

P a rticip a n t(3 6 % )

P la ce boC a lc iu m a n d D

C a lc iu m a n d D

D e ta ils o fa p la ce bo

co n tro lle d T ria l

R e fusedto

P a rticip a n t(2 6 % )

C a lc iu m a n d D N o tre a tm e nt

D e ta ils o fa n o pen

co n tro lle d tria l.

M e n & W o m en7 0 + w ith a rece n t

fra c tu reR a nd o m ised

Diagram of OPEN trial

R e fusedT o

P a rticip a n t(3 6 % )

P la ce boC a lc iu m a n d D

C a lc iu m a n d D

D e ta ils o fa p la ce bo

co n tro lle d T ria l

R e fusedto

P a rticip a n t(2 6 % )

C a lc iu m a n d D N o tre a tm e nt

D e ta ils o fa n o pen

co n tro lle d tria l.

M e n & W o m en7 0 + w ith a rece n t

fra c tu reR a nd o m ised

Trial retention To avoid bias it is CRUCIAL that after

randomisation as many participants are retained within the trial as possible.

Using an OPEN design we might expect that those randomised to no treatment might drop out to seek treatment elsewhere.

Important to test this – fortunately, Avenell and colleagues did so.

Retention Rates: Placebo vs Open

0%

20%

40%

60%

80%

100%

At baseline At 4 months At 8 months At 12 months

OPEN (All)

RECORD (All)

Overall retention This seemed to be better in the

open trial compared with the placebo study.

BUT was active group being retained more often than the open treatment group?

Retention Rates by Group

0%

20%

40%

60%

80%

100%

At baseline At 4 months At 8 months At 12 months

OPEN (Active)

OPEN (No tablets)

RECORD (Active)RECORD (Placebo)

The OPEN trial again Interestingly the drop out rate in

the OPEN trial is lowest in the NO treatment group. This is CONTRARY to expectations as we assumed this would be the highest.

LESSON – Trialists need to test trial ideas in RANDOMISED TRIALS.

Another ‘OPEN’ trial Hemminki using similar methods as

Avenell et al, randomised 4,295 women to take part in a placebo controlled trial of HRT or an open trial of HRT.

Those randomised to the open approach 48% were recruited compared with 37% for placebo group. The difference 11% was significant and similar to the Avenell study.

Hemminki et al. J Clin Epidemiol 2004;57:1237.

Placebos - conclusions For phase 1 & II trials placebos are

useful to make sure there is a ‘true’ physiological or psychological effect. Their use, however, is not problem free.

They are probably used TOO frequently for phase III or IV trials.

Summary Placebos are effective at removing

some forms of bias. Care needs to be taken that

placebos are developed correctly. Placebos are not PRAGMATIC and

can lead to results that are not replicable in normal care.